From d33fd00cfe47dc4685614daf9248a2d813147399 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Mon, 7 Apr 2025 13:18:07 +0300 Subject: [PATCH] cuda : fix HIP and MUSA BF16 (llama/0) ggml-ci --- ggml/src/ggml-cuda/convert.cu | 8 +------- ggml/src/ggml-cuda/vendors/hip.h | 1 + ggml/src/ggml-cuda/vendors/musa.h | 1 + 3 files changed, 3 insertions(+), 7 deletions(-) diff --git a/ggml/src/ggml-cuda/convert.cu b/ggml/src/ggml-cuda/convert.cu index 2681c42e..a224ec0e 100644 --- a/ggml/src/ggml-cuda/convert.cu +++ b/ggml/src/ggml-cuda/convert.cu @@ -579,13 +579,7 @@ static __global__ void convert_unary(const void * __restrict__ vx, dst_t * __res const src_t * x = (const src_t *) vx; - if constexpr (std::is_same_v) { - y[i] = __bfloat162float(x[i]); - } else if constexpr (std::is_same_v && std::is_same_v) { - y[i] = (float)x[i]; - } else { - y[i] = x[i]; - } + y[i] = float(x[i]); } template diff --git a/ggml/src/ggml-cuda/vendors/hip.h b/ggml/src/ggml-cuda/vendors/hip.h index 3983ce5b..420b41b8 100644 --- a/ggml/src/ggml-cuda/vendors/hip.h +++ b/ggml/src/ggml-cuda/vendors/hip.h @@ -20,6 +20,7 @@ #define CUBLAS_STATUS_SUCCESS HIPBLAS_STATUS_SUCCESS #define CUBLAS_TF32_TENSOR_OP_MATH 0 #define CUDA_R_16F HIPBLAS_R_16F +#define CUDA_R_16BF HIPBLAS_R_16B #define CUDA_R_32F HIPBLAS_R_32F #define CU_DEVICE_ATTRIBUTE_VIRTUAL_MEMORY_MANAGEMENT_SUPPORTED hipDeviceAttributeVirtualMemoryManagementSupported #define CU_MEM_ALLOC_GRANULARITY_RECOMMENDED hipMemAllocationGranularityRecommended diff --git a/ggml/src/ggml-cuda/vendors/musa.h b/ggml/src/ggml-cuda/vendors/musa.h index f2d55796..937779a9 100644 --- a/ggml/src/ggml-cuda/vendors/musa.h +++ b/ggml/src/ggml-cuda/vendors/musa.h @@ -15,6 +15,7 @@ #define CUBLAS_STATUS_SUCCESS MUBLAS_STATUS_SUCCESS #define CUBLAS_TF32_TENSOR_OP_MATH MUBLAS_MATH_MODE_DEFAULT #define CUDA_R_16F MUSA_R_16F +#define CUDA_R_16BF MUSA_R_16BF #define CUDA_R_32F MUSA_R_32F #define cublasComputeType_t cudaDataType_t #define cublasCreate mublasCreate